---
name: kb-article-creation
description: Automatically generate knowledge base articles from resolved support tickets, including content extraction, quality review, approval workflow, and publishing. Use when automating KB content creation, extracting solutions from tickets, designing content review workflows, or maintaining fresh knowledge base coverage. Triggers on phrases like "KB article creation", "automated knowledge base", "article from ticket", "solution extraction", "content automation", "knowledge capture", "KB generation", "article drafting".
---

# Automated KB Article Creation from Tickets

Automatically generate knowledge base articles from resolved support tickets.

## Workflow

### Automated Article Generation Process

Trigger: Weekly batch processing; high-volume resolved tickets; knowledge gap detected:

1. **Ticket identification**: Scan resolved tickets from last 7 days; filter for: resolution provided, CSAT ≥ 4.0, no escalation, repeatable solution (not account-specific); exclude: billing-only, security-sensitive, one-off fixes.
2. **Solution extraction**: Parse agent response to extract core solution (numbered steps, configuration details, error codes); identify Q&A pattern (customer question → agent answer); extract relevant screenshots and links.
3. **Article drafting**: Generate structured article using template — Title (question format), Summary (1–2 sentences), Prerequisites, Steps (numbered), Troubleshooting (common issues), Related articles; format in Markdown/HTML.
4. **Similarity check**: Compare draft against existing KB articles; if similarity > 80%, flag as "update existing article" instead of "create new"; if 50–80%, suggest merge; if < 50%, proceed as new article.
5. **Quality review queue**: Route draft to content reviewer (support lead or dedicated writer); reviewer edits for clarity, accuracy, completeness; adds screenshots if needed; assigns category and tags.
6. **Approval workflow**: Reviewed article sent to subject matter expert (SME) for accuracy verification; SME approves, requests changes, or rejects; track approval SLA (48 hours).
7. **Publishing**: Approved article published to knowledge base; tagged with source ticket IDs; linked from relevant tickets; indexed for search; notified to subscribers.
8. **Feedback loop**: Track article performance (views, helpful votes, time on page); flag low-performing articles for rewrite; retire outdated articles quarterly.

### Article Generation Template

```
KB ARTICLE TEMPLATE — AUTO-GENERATED DRAFT
=============================================

[Article Title — question format]
"How do I [resolve X issue]?"

[Summary]
1–2 sentence description of the problem and solution.

[Last updated: [Date] | Category: [Category] | Tags: [tag1, tag2, tag3]]

---

## Problem Description

[What the customer is experiencing]
[Error message if applicable]
[When this issue occurs]

## Prerequisites

- [Requirement 1 — e.g., Admin access]
- [Requirement 2 — e.g., Product version 2.0+]
- [Requirement 3 — e.g., API access enabled]

## Solution

### Step 1: [Action]
[Detailed instruction with screenshots]
[Expected result]

### Step 2: [Action]
[Detailed instruction]
[Expected result]

### Step 3: [Action]
[Detailed instruction]
[Expected result]

## Verification

[How to confirm the issue is resolved]
[Expected behavior after fix]

## Troubleshooting

### If [issue] persists:
- [Check 1]
- [Check 2]
- [Alternative solution]

### Common variations:
- [Variation 1]: [Specific fix]
- [Variation 2]: [Specific fix]

## Related Articles

- [Link to related article 1]
- [Link to related article 2]

## Was this article helpful?
[ ] Yes   [ ] No

---

SOURCE: Generated from tickets #[12345, 12346, 12347]
REVIEWER: [Name]
APPROVED BY: [SME Name]
```

### Article Quality Standards

```
KB ARTICLE QUALITY CHECKLIST
===============================

Clarity (Score 1–5):
  □ Title clearly describes the problem (not vague)
  □ Steps are numbered and sequential
  □ Each step is one action (not combined)
  □ Language is simple (8th-grade reading level)
  □ No jargon without definition

Completeness (Score 1–5):
  □ All prerequisites listed
  □ All steps required to resolve issue included
  □ Verification step included ("how to confirm it works")
  □ Troubleshooting section covers common issues
  □ Screenshots attached for complex steps

Accuracy (Score 1–5):
  □ Solution verified against current product version
  □ Error messages/paths match actual UI
  □ Links to related articles are correct
  □ No outdated information from old product versions
  □ SME-approved for technical accuracy

Structure (Score 1–5):
  □ Follows standard template
  □ Proper headings hierarchy (H2, H3)
  □ Code blocks formatted correctly
  □ Screenshots labeled and positioned near relevant step
  □ Related articles linked (2–4 articles)

SEO (Score 1–5):
  □ Title contains primary search term
  □ First paragraph includes key terms
  □ Meta description written (155 characters)
  □ URL slug is clean and descriptive
  □ Internal links to related articles

Minimum Score: 20/25 (4.0 average) to publish
Articles below threshold: Send back for revision
```

## Edge Cases

- **Account-specific solutions** (fix only works for particular customer configuration):
  - Detection: Solution references specific account ID, custom integration, unique configuration
  - Handling: Flag as "not generalizable"; do not create public KB article; save as internal note
  - Alternative: Create "internal KB" article for agents; generalize if pattern repeats with other customers
  - Review: Quarterly review of internal KB for generalizable patterns

- **Sensitive information in tickets** (customer PII, security credentials, proprietary data):
  - Protection: PII redaction before article drafting (names, emails, phone numbers, account IDs); credential masking (API keys, passwords, tokens)
  - Detection: Automated PII scan (email pattern, phone pattern, SSN pattern); manual review by content writer
  - Prevention: Never include customer-specific data in public articles; use generic examples ("e.g., company@example.com")
  - Compliance: GDPR/CCPA compliance — no personal data in public KB

- **Rapidly changing product** (article outdated before review completes):
  - Detection: Product release notes check during review; version tags on articles; "Last verified" date
  - Prevention: Tie article updates to product release cycle; automated flag when product version changes
  - Communication: Show "This article applies to version X" prominently; alert if newer version exists
  - SLA: Update articles within 72 hours of product release affecting covered features

- **Low-quality source tickets** (agent response is vague, incomplete, or unhelpful):
  - Detection: CSAT < 4.0; response < 50 words; no numbered steps; follow-up tickets on same issue
  - Handling: Exclude from automated drafting; flag for content writer to research independently
  - Prevention: Agent training on solution documentation; encourage detailed, structured responses
  - Quality gate: Minimum criteria before extraction — CSAT ≥ 4.0, response ≥ 100 words, numbered steps present

- **Duplicate/near-duplicate articles** (multiple articles covering same topic):
  - Detection: Semantic similarity check (> 80% similarity); title comparison; category overlap
  - Handling: Merge into single authoritative article; redirect old URLs to new article; update internal links
  - Prevention: Similarity check before publishing; "existing articles" suggestion during drafting
  - Audit: Quarterly article deduplication review; retire redundant articles; maintain single source of truth

## Integration Points

- **Help desk**: Zendesk, Freshdesk, Intercom — ticket data extraction, resolution parsing
- **Knowledge base**: Zendesk Guide, Confluence, Notion, Helpjuice — article publishing, version control
- **NLP services**: OpenAI, Google Cloud NL, Azure AI — content extraction, summarization, classification
- **Similarity search**: Elasticsearch, Pinecone, Weaviate — duplicate detection, semantic search
- **Workflow automation**: Zapier, Make, custom — article drafting pipeline, review routing
- **CMS**: WordPress, Contentful — content management, publishing, SEO
- **Analytics**: Google Analytics, custom — article views, helpful votes, search performance
- **Screenshot tools**: Gyazo, CleanShot X, custom — automatic screenshot capture from product
- **Version control**: Git, GitHub — article versioning, change tracking
- **Data warehouse**: Snowflake, BigQuery — article performance analytics, content gap analysis